Diagnosing epilepsy using entropy measures and embedding parameters of EEG signals
Authors:
Fatemeh Valipour,
Zahra Valipour,
Mani Garousi,
Ali Khadem
Abstract:
Epilepsy is a neurological disorder that affects normal neural activity. These electrical activities can be recorded as signals containing information about the brain known as Electroencephalography (EEG) signals. Analysis of the EEG signals by individuals for epilepsy diagnosis is subjective and time-consuming. So, an automatic classification system with high detection accuracy is required to ove…
▽ More
Epilepsy is a neurological disorder that affects normal neural activity. These electrical activities can be recorded as signals containing information about the brain known as Electroencephalography (EEG) signals. Analysis of the EEG signals by individuals for epilepsy diagnosis is subjective and time-consuming. So, an automatic classification system with high detection accuracy is required to overcome possible errors. In this study, the discrete wavelet transform has been applied to EEG signals. Then, entropy measures and embedding parameters have been extracted. These features have been investigated individually to find the most discriminating ones. The significance level of each feature was evaluated by statistical analysis. Consequently, LDA and SVM algorithms have been employed to categorize the EEG signals. The results have indicated that the features of Embedding parameters, PermutationEntropy, FuzzyEntropy, SampleEntropy, NormEntropy, SureEntropy, LogEntropy, and ThresholdEntropy have the potential to discriminate epileptic patients from healthy subjects significantly. Also, SVM classifier has achieved the highest classification accuracy. In this study, we could find effective embedding-based and entropy-based features as appropriate single measures for identifying abnormal activities that can efficiently discriminate the EEG signals of epileptics from healthy individuals. According to the results, they can be used for automatic classification of epileptic EEG signals that are difficult to examine visually.
△ Less
Submitted 18 January, 2024; v1 submitted 14 January, 2024;
originally announced January 2024.
